More shonky science

Over at Crooked Timber, Daniel Davies demolishes the latest effort by junk science writer Steve Milloy (on diet and diabetes), and admonishes Todd Zywicki of the Volokh Conspiracy for giving uncritical credence to Milloy, who has a track record of bogus work going back at least a decade, to his work for the tobacco companies, trying to cast doubt on the link between smoking and lung cancer. Of course, this track record doesn’t eliminate the need to demolish Milloy afresh each time he pops up. Still if the name didn’t ring alarm bells for Zywicki as soon as he saw it, it’s clear he hasn’t been paying attention.

Devaluation of scientific work in favor of partisan hacks like Milloy has become standard practice on the political right and is a significant subtext in the current campaign against the Greens. Many of those making these attacks are drawing their talking points from the Institute of Public Affairs, a body which has supported global warming contrarianism, denied the link between passive smoking and cancer, and attacked scientists working on the Murray-Darling basin as “environmental activists masquerading as scientists”. It’s pretty clear who the wackos are in this debate.

Meanwhile, in email Tim Lambert advises me that Ross McKitrick has conceded[1] that he erroneously used latitude in degrees when it should have been radians in his work on causes of global warming.

In his defence, McKitrick makes the point that the detection of this error

supports our contention that empirical researchers ought to publish both their data and code in easily accessible format, even if the example in this case comes at my expense. Empirical work on large data bases is easily prone to glitches in programming or data handling. This is inevitable. So where novel and possibly controversial results are obtained it is important to ensure complete transparency in the supporting materials.

McKitrick is right on this point, and has succeeded in demonstrating that the documentation provided by Mann et al for their ‘hockey stick’ paper was inadequate, though not in showing that this made any material difference to results.

On the other hand, a trawl back through the files makes it pretty clear that this error was not exactly an innocent mistake. It seems pretty clear that McKitrick tried some regressions with (absolute) latitude as the explanatory variable, didn’t like the results he got and switched to the cosine (note that, if you were starting here, you wouldn’t need to take the absolute value, since cosine is a symmetric function). Because of the degrees-radians mistake, this variable came out insignificant, as desired, and McKitrick didn’t do the checks that would have revealed the error. Asymmetric error-checking is a standard problem with cherry picking, as illustrated by the work of John Lock Lott.

fn1. In climatesceptics, a members-only Yahoo group, so I can’t give a link.

12 thoughts on “More shonky science

  1. Sorry, John, but could you explain “Asymmetric error-checking”? I assume there is a “symmetric error-checking” as well.

  2. Peter, what I mean here is that, if the results come out in a way that you don’t expect or want, you check them carefully for possible errors. If they come out the way you want, you take them and run with them.

  3. The rhetorical counterpart to “asymmetric error-checking” is what I call ‘selective fastidiousness’. This is where one imposes an impossibly high burden of proof on one’s opponents and a thoroughly breezy one on oneself. This is standard operating procedure in bureaucratic argumentation (though I suspect we all suffer from it I to some extent). A touchstone of its worst excesses is the use of words like ‘could’. Some policy ‘could’ cause disruption, or inefficiency, not ‘will’ or even ‘is likely to’ but ‘could’. When you run into this kind of thing, you know you’re not really engaged in a bona fide conversation. The moon ‘could’ be made of cream cheese. But it probably isn’t!

  4. A good marker of selective fastidious is frequent use of the term “flawed”. It’s almost impossible to deny (who could honestly describe their work as “flawless”), but carries the imputation that the work in question is valueless

  5. Small correction: What he wanted to show was that economic factors were a strong influence. Taking the cosine and screwing up the calculation makes the economic effects stronger, which is why he did it. The only honest justification for taking the cos would be if it improved the fit of the model, but it actually makes the fit much worse.

  6. I think you mean “John Lott” rather than “John Lock” but in the interests of having a non-dogmatic Bayesian prior I’m putting small nonzero probability on “John Locke” too.

  7. Nic’s right about the weaselly language of ‘could’. But I’ve seen much worse examples of dishonest arguments – memorably part of a draft cabinet submission which began, in support of the proposal being put, with the immortal phrase “Research will find that …”

  8. These are classic cases of confirmation bias / file draw problem. It’s used by many of those with political biases against regulation of business on matters concerning the environment. From Global Warming, Acid Rain, Ozone Hole and Biodiversity Loss they search for fringe scientists going against mainstream science or them-selves misrepresenting or plain lying about the studies of mainstream science or have serious conflicts of interest. Many have philosophical connections with economic rationalists or have conservative points of view.

    Andrew Bolt misrepresented work of social historians to try and connect The Greens and Nazism and told us on the Insiders -using the work of The Institute of Public Affairs- that the Murray is the best it has been in 20 years.

    You can argue to the cows come home on differences on values and the like but you should let the science speak for itself.

Comments are closed.